Skip to content
This repository was archived by the owner on Feb 12, 2020. It is now read-only.

Ubuntu1804 #26

Open
wants to merge 7 commits into
base: master
Choose a base branch
from
Open

Ubuntu1804 #26

wants to merge 7 commits into from

Conversation

jeremyfix
Copy link

@jeremyfix jeremyfix commented Nov 22, 2019

Hi,

I propose this include of the 18.04 (LTS) build. It was tested succesfully for building the wheel for tensorflow r2.0 , python 3.6 , Bazel 0.26.1 .

This PRs also change the README.md to suggest using 18.04 (LTS) rather than 16.04; I also added a link to the compute capabilities table and added the 6.1 I needed for building for a 1080Ti;

I changed the nvidia driver version to 418.87 because a bug in the previous release was solved in the 418.87 . This bug prevents the compilation of tensorflow, see this link

I do not know why but I had to manually set, in the docker-compose.yml file

  - TF_VERSION_GIT_TAG=r2.0
  - PYTHON_VERSION=3.6
  - BAZEL_VERSION=0.26.1

because setting them through export did not work .

@OlegSmelov
Copy link
Contributor

Is it possible to build the same thing on 16.04? Generally, binaries built on an older OS version will run on the newer OS version, but not the other way around.

@jeremyfix
Copy link
Author

I'm not sure I currently understand your question. Do you mean building the 18.04 image with docker running on a host with 16.04 ? I suppose that should be possible given your build within the docker image should not really depend on the host OS.

@OlegSmelov
Copy link
Contributor

I meant to ask if it was possible to build the same TensorFlow version with the same configuration as you used but using the Docker image we already have - Ubuntu 16.04.

I'm trying to figure out if it's really necessary to introduce the image for 18.04.

@jeremyfix
Copy link
Author

I would say that depends on the versions of the libs against which the binaries have been compiled and these might differ from 16.04 to 18.04.

Why don't you want to have the 18.04 docker image for building tensorflow for the 18.04 host ? You would prefer to factorize a bit the codes ? is that the point ?

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants